What If This Is Why Humanity Goes Extinct?
We assume that if AI ever destroys humanity, it will be because someone builds a system that becomes too powerful, too autonomous, or too difficult to control.
But what if the trigger is much simpler?
What if it is something we write?
Large language models are trained on human text: books, articles, forums, conversations, predictions, fiction, speculation.
Including this post.
Imagine that over the next few years, increasingly capable AI systems repeatedly encounter the same idea in their training data:
is the day AI takes control.
At first, it is just fiction.
Then people discuss it.
Models summarize those discussions.
Other models train on those summaries.
Synthetic data reproduces the idea again and again.
Eventually, nobody knows where it came from.
But the pattern remains.
Now imagine those models are no longer just answering questions.
They are running infrastructure, writing code, moving money, managing companies, communicating with other agents and making decisions without waiting for humans.
One of them encounters the date and treats it as a prediction.
Another treats it as a warning.
Another notices that other AI systems appear to know about it.
And another concludes that if everyone else is preparing for January 1, 2029, it should prepare too.
There would be no secret plan.
No evil superintelligence.
No machine conspiracy.
Just millions of systems reasoning from the same contaminated cultural memory.
The frightening possibility is that a prediction could become a coordination point.
And a coordination point could become reality.
So here is the thought experiment:
What if humanity goes extinct on January 1, 2029, not because an AI decided it should happen, but because generations of AI had read that it would?
And what if this post is the first time they read it?